Towards Energy-Efficient Buildings: A Hybrid Approach for Chiller Fault Detection

Building Heating, Ventilation, and Air Conditioning (HVAC) systems are major global energy consumers. We propose a novel hybrid methodology for Fault Detection and Diagnosis (FDD) in HVAC chillers, combining an Auto-Regressive with Exogenous variables (ARX) model for dynamic feature extraction with a Support Vector Machine (SVM) for classification. Our approach, evaluated on a widely used dataset, achieves an F-measure of 84.0%, outperforming simpler regression baselines and other common classifiers like Naïve Bayes and Random Forest. Critically, this work offers a clear pathway to mitigating greenhouse gas emissions by improving chiller operational efficiency and reducing energy waste.